Xushaohong commented on PR #4052:
URL: https://github.com/apache/ozone/pull/4052#issuecomment-1347819690

   > How does this work in presence of HA failovers and OM leader changing? 
   
   The whole install of snapshot may take a few times, which involves one full 
snaphsnapshot and the other incremental snapshots.
   In my design, the follower only installs the Incremental snapshot from the 
same leader. This is protected by an AtomicReference lastLeader and the 
function RDBSnapshotProvider#checkLeaderConsistent, which will check the 
previous leader if is the same one and also check the candidate DB dir(which 
stores the snapshot downloaded at the follower side).
   
   > For the issue of RocksDB compaction that can throw off SST file comparison.
   
   Here my SST file comparison is a static comparison which is just a file name 
comparision. Since the RDB checkpoint from the same DB ensures each SST file 
with the same name is totally same. This might be a little tricky, and I have 
the UT for this.
   
   > We have a Snapshot feature that can help with something like this. It does 
take care of OM HA as well as rocksDB compaction. This can all be simplified if 
you can leverage the Snapshot feature. We can also do a zoom call during common 
time window to go over details.
   
   Last week, I have quickly scanned the Snapshot Feature branch, and some 
contents could be unified with mine. I agree with doing a Zoom call, how about 
you scheduling a time (GMT+8), and let's have a sync up? @prashantpogde 


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to